CN101335002A - Method and apparatus for audio decoding - Google Patents

Abstract

The invention provides a decoding method of an audio signal. The encoding method includes the following steps: when the audio signal corresponding to the received coding stream switched from the relatively wide bandwidth to the relatively narrow bandwidth, low band signal component of the audio signal is obtained and high band information is extended; the time-varying fade-out treatment is carried out to extended high band information to obtain processed high band information component; the processed high band information component and the low band signal component are combined. By the method provided by the embodiment of the invention, when the audio signal switches from the wide band to narrow band, a series of treatments, such as bandwidth detection, manual band spreading, time-varying fade-out treatment and bandwidth combination are available to realize smoothing switch from a wide band signal to a narrow band signal, thus achieving a more comfortable auditory feeling of human ears.

Description

A kind of method and apparatus of audio decoder

Technical field

The present invention relates to the voice communication field, relate in particular to a kind of method and apparatus of audio decoder.

Background technology

G.729.1 be ITU (International Telecommunication Union, International Telecommunications Union (ITU)) the encoding and decoding speech standard of new generation of up-to-date issue, the characteristics of this embedded speech encoding and decoding standard maximum are the characteristics with hierarchical coding, can provide range of code rates in the arrowband of 8kb/s～32kb/s the audio quality to the broadband, permission is in transmission course, abandon outer code stream according to channel conditions, have good channel self-adapting.

In standard G.729.1, reach graded properties by code stream being configured to Embedded hierarchy, but be a kind of multi-rate speech codec of novel embedded layering.Be input as the superframe of 20ms, when sampling rate is 16000Hz, frame length is 320 points.Fig. 1 is each layer coder system chart G.729.1, and the detailed process of audio coder ﹠ decoder (codec) coding is: input signal s _WB(n) at first pass through QMF (QuadratureMirror Filterbank, quadrature mirror filter) filtering and be divided into (H ₁(z), H ₂(z)) two subbands, low subband signal s _LB ^Qmf(n) Hi-pass filter through the 50Hz cutoff frequency carries out pre-service, output signal s _LB(n) use arrowband embedded type C ELP (Code-Excited Linear-Prediction, the Code Excited Linear Prediction) scrambler of 8kb/s～12kb/s to encode s _LB(n) the local composite signal of celp coder and under the 12Kb/s code check Between difference signal d _LB(n) through perceptual weighting filtering (W _LB(z)) the signal d after _LB ^w(n) transform to frequency domain by MDCT (Modified Discrete Cosine Transform, the discrete cosine transform of correction).Weighting filter W _LB(z) comprise gain compensation, be used for keeping wave filter output d _LB ^w(n) with high subband input signal S _HB(n) the spectrum continuity between.Difference signal after the weighting will transform in the frequency domain.

High subband component is multiplied by (1) ⁿCarry out spectral inversion signal s afterwards _HB ^Fold(n) be that the low-pass filter of 3000HZ carries out pre-service, filtered signal s by cutoff frequency _HB(n) use TDBWE (Time-Domain BandWidth Extension, time domain bandwidth extended coding algorithm) scrambler to encode.Enter the s of TDAC (Time Domain Alias Cancellation, the time domain aliasing is eliminated) coding module _HB(n) also to use MDCT to transform on the frequency domain earlier.

Two groups of MDCT coefficient D _LB ^w(k) and S _HB(k) use the TDAC encryption algorithm to encode at last.In addition, also have some parameters to transmit, the mistake that causes when in transmission, frame losing occurring in order to improve with FEC (Frame Erasure Concealment, frame error concealment) scrambler.

Fig. 2 is each layer decoder system chart G.729.1, and the real work pattern of demoder also is equivalent to by the code check decision that receives by the code stream number of plies decision that receives.As follows according to each situation division of different code checks that receiving end receives:

If 1 code check that receives is 8kb/s or 12kb/s (promptly only receiving ground floor or preceding two-layer): ground floor or preceding two-layer code stream are decoded by embedded type C ELP demoder, obtain decoded signal

Carrying out back filtering again obtains

Through entering the broadband signal of the synthetic 16kHz of QMF bank of filters after the high-pass filtering, wherein the high-band component of signal puts 0.

If 2 code checks that receive are 14kb/s (promptly receiving three first layers): except the CELP decoder decode went out the arrowband component, the TDBWE demoder also decoded the high-band component of signal

Right

Carry out the MDCT conversion, 3000Hz in the high subband component spectrum above (corresponding in the 16kHz sampling rate more than the 7000Hz) frequency component is put 0, carry out contrary MDCT conversion then, after the superposition and carry out spectrum inversion, the low strap component that in the QMF bank of filters, solves then with the CELP demoder

Synthetic together sampling rate is the broadband signal of 16kHz.

If 3 code streams that receive the above speed of 14kb/s (corresponding to preceding four layers or more multi-layered): except the CELP decoder decode goes out to hang down the subband component

The TDBWE decoder decode goes out high subband component

In addition, also will use the TDAC decoder decode to go out to hang down subband weighted difference signal and high subband enhancing signal, the full range band signal is strengthened, finally also synthesizing sampling rate in the QMF bank of filters is the broadband signal of 16kHz.

The code stream of G729.1 has hierarchy, and permission transmittability according to channel in the process of transmission abandons outer code stream from outside to inside, to reach the self-adaptation to channel conditions.By in the arthmetic statement of encoding and decoding as can be seen, if when channel capacity changes in time faster, demoder may the time and receive that arrowband code stream (being equal to or less than 12kb/s), the signal that decode this moment only comprise the following component of 4000Hz; The time and receive broadband code stream (equal or be higher than 14kb/s), the signal that decode this moment then comprises the broadband signal of 0～7000Hz.The unexpected variation of this bandwidth, we are referred to as bandwidth switches, because high low strap is also inequality to the contribution of people's ear impression, therefore so frequent switching brings tangible discomfort can for people's ear.Especially, when frequent appearance by the broadband during to the switching of arrowband, people's ear can obviously feel continually the sound heard from clear and melodious transition for dull, therefore need to use a kind of technology alleviate the discomfort that this frequent switching brings human auditory system.

Summary of the invention

The embodiment of the invention provides a kind of method and apparatus of audio decoder, and purpose is the comfort level of people's ear impression in the time of will improving the speech signal bandwidth switching.

For achieving the above object, the embodiment of the invention provides a kind of method of audio decoder, may further comprise the steps:

When the encoding code stream corresponding audio signal that receives by wider bandwidth when switching than narrow bandwidth, obtain the low strap component of signal of described sound signal and expand high-band information;

When being carried out, the described high-band information that expands becomes gradually to go out to handle the high-band component of signal after obtaining handling;

High-band component of signal after the described processing and the described low strap component of signal of obtaining are synthesized.

The embodiment of the invention also provides a kind of device of audio decoder, comprising:

Acquiring unit, be used for when the encoding code stream corresponding audio signal that receives by wider bandwidth when switching than narrow bandwidth, obtain the low strap component of signal of described sound signal, and send to expanding element;

Expanding element is used for described low strap component of signal is expanded high-band information, and becomes when the described high-band information that expands sent to and gradually go out processing unit;

In time, becomes and gradually to go out processing unit, becomes gradually to go out to handle when being used for the described high-band information that expands carried out, and the high-band component of signal after obtaining handling, and the high-band component of signal after the described processing sent to synthesis unit;

Synthesis unit, the low strap component of signal that high-band component of signal after the described processing that is used for receiving and described acquiring unit obtain is synthesized.

The embodiment of the invention has following beneficial effect:

The method that provides by the embodiment of the invention, when sound signal taken place the broadband to the switching of arrowband the time, can utilize bandwidth detection, artificial band spread, time to become gradually to go out to handle and a series of processing such as bandwidth is synthetic, can carry out the transition to narrow band signal from broadband signal smoothly to realize switching, thereby make people's ear obtain comparatively comfortable auditory perception.

Description of drawings

Fig. 1 is an encoder system block diagram G.729.1 in the prior art;

Fig. 2 is a decoder system block diagram G.729.1 in the prior art;

Fig. 3 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention one;

Fig. 4 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention two;

Fig. 5 is a time-varying gain factor variations curve in the embodiment of the invention two;

Fig. 6 is that the time varying filter limit changes synoptic diagram in the embodiment of the invention two;

Fig. 7 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention three;

Fig. 8 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention four;

Fig. 9 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention five;

Figure 10 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention six;

Figure 11 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention seven;

Figure 12 is the coding/decoding method process flow diagram of a kind of sound signal of the embodiment of the invention eight;

Figure 13 is the decoding device synoptic diagram of a kind of sound signal of the embodiment of the invention nine.

Embodiment

Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail.

In the embodiments of the invention one, a kind of coding/decoding method of sound signal as shown in Figure 3, concrete steps are as follows:

Step S301, determine the frame structure of the encoding code stream that receives.

Step S302, detect encoding code stream corresponding audio signal according to the frame structure of encoding code stream and whether taken place by wider bandwidth to switching than narrow bandwidth, if switching has taken place, change step S303, otherwise, encoding code stream decoded and export sound signal after the reconstruction according to normal decoding process.

In general, in the encoding and decoding speech field, narrow band signal is meant the signal of frequency band 0～4000Hz, and broadband signal is meant the signal of frequency band at 0～8000Hz, and ultra-broadband signal is meant the signal of frequency band at 0～16000Hz.The signal of frequency band broad can be decomposed into low strap component of signal and high-band component of signal again.Certainly, the definition here is in general sense, can be not limited thereto during practical application.For convenience of description, the high-band component of signal in the embodiments of the invention is meant the part of switching the back and increasing with respect to bandwidth before switching, and the narrow band signal component is before switching and the portions of bandwidth that all has of the sound signal after switching.For example, when the signal of frequency band 0～4000Hz switched, the low strap component of signal referred to the signal of 0～4000Hz to frequency band at the signal of 0～8000Hz, and the high-band component of signal refers to the signal of 4000～8000Hz.

Step S303, detect encoding code stream corresponding audio signal by wider bandwidth when switching than narrow bandwidth, utilize the low strap coding parameter receive to decode the low strap component of signal.

In the embodiment of the invention, as long as the broader bandwidth before the bandwidth ratio before switching is switched can be used the scheme of the embodiment of the invention, and switch to the arrowband in the broadband that is not confined in general sense.

Step S304, utilize artificial band spreading technique, the low strap component of signal is expanded high-band information.

Concrete, described high-band information can be high-band component of signal or high-band coding parameter.Utilize artificial band spreading technique, encoding code stream corresponding audio signal by wider bandwidth in the initial a period of time of switching than narrow bandwidth, the method that the low strap component of signal is expanded high-band information can be following two kinds: utilize the high-band coding parameter that receives before switching that the low strap component of signal is expanded high-band information; Or the low strap component of signal of utilizing the current audio frame in switching back to decode expands high-band information.

The high-band coding parameter that receives before utilize switching is specially the method that the low strap component of signal expands high-band information: the high-band coding parameter that buffer memory receives before switching (TDBWE (Time-DomainBandwidth Extension for example, the expansion of time domain bandwidth) the MDCT coefficient in time domain, frequency domain envelope or the TDAC encryption algorithm in the encryption algorithm), after switching, utilize extrapolation to estimate the high-band coding parameter of current audio frame, further, can expand the high-band component of signal according to the corresponding wideband decoded algorithm of high-band coding parameter utilization.

The method that the low strap component of signal of utilizing the current audio frame in switching back to decode expands high-band information is specially: the low strap component of signal that current audio frame after switching decodes is carried out FFT conversion (Fast FourierTransform, fast fourier transform), FFT coefficient to the low strap component of signal carries out continuation and shaping in the FFT territory then, FFT coefficient after the shaping carries out anti-FFT conversion again and can expand the high-band component of signal as the FFT coefficient of high-band information.Certainly, a kind of method computation complexity will be well below a kind of method in back before using, and a kind of method is that present invention is described for example before all adopting in following examples.

Step S305, when being carried out, the high-band information that expands becomes gradually to go out to handle.

Concrete, utilize artificial band spreading technique to expand high-band information after, do not carry out QMF filtering with high-band information and low strap component of signal synthesized wideband signal, but become gradually to go out to handle when the high-band information that expands carried out.Gradually go out to handle is sound signal by wider bandwidth to than the narrow bandwidth transition.Becoming the method that gradually goes out to handle when high-band information is carried out specifically comprises: become gradually to go out to handle when change gradually goes out processing and mixing during separation.

Become during separation gradually to go out to handle and be specially: method one, utilize the time domain gain factor that the high-band information that expands is carried out the time domain shaping, further can utilize time-variable filtering to carry out frequency-domain shaping the high-band information after the time domain shaping; Or method two, utilize time-variable filtering to carry out frequency-domain shaping to the high-band information that expands, further can utilize the time domain gain factor to carry out the time domain shaping to the high-band information after the frequency-domain shaping.

Become during mixing gradually to go out to handle and be specially: method three, utilize frequency domain high-band parameter time varying method of weighting that the high-band information that expands is carried out frequency-domain shaping, become the spectrum envelope that gradually goes out when obtaining, the high-band component of signal after decoding obtains handling; Or method four, the high-band information that expands is divided subband, and each sub-band coding parameter is carried out the weighting of frequency domain high-band parameter time varying, become the spectrum envelope that gradually goes out when obtaining, the high-band component of signal after decoding obtains handling.

Step S306, the high-band component of signal after will handling and the low strap component of signal that decodes are synthesized.

In the above-mentioned steps, demoder has the disposal route that becomes when multiple gradually to go out to the high-band information that expands, and becomes the disposal route that gradually goes out when different below and is described in detail in the mode of specific embodiment respectively.

In following examples, the encoding code stream that demoder receives can be voice segments, and voice segments is meant one section speech frame that demoder receives continuously, and speech frame can be the full-speed voice frame, also can be the several layers of full-speed voice frame.The encoding code stream that demoder receives also can be the noise section, and the noise section is meant one section noise frame that demoder receives continuously, and noise frame can be the full rate noise frame, also can be the several layers of full rate noise frame.

In the embodiments of the invention two, the encoding code stream that receives with demoder is a voice segments, and in time, becomes and gradually to go out to handle employing method one, promptly utilize the time domain gain factor that the high-band information that expands is carried out the time domain shaping, further can utilize time-variable filtering to carry out frequency-domain shaping and be example the high-band information after the time domain shaping, a kind of coding/decoding method of sound signal as shown in Figure 4, concrete steps are as follows:

Step S401, demoder receive the encoding code stream that scrambler sends, and determine the frame structure of the encoding code stream that receives.

Concrete, scrambler adopts the flow process of system shown in Figure 1 block diagram to coding audio signal, encoding code stream is sent to demoder, demoder receives encoding code stream, if encoding code stream corresponding audio signal does not take place by the switching of broadband to the arrowband, demoder adopts the flow process of system chart shown in Figure 2 that the encoding code stream that receives is carried out normal decoder, does not repeat them here.The encoding code stream that demoder receives is a voice segments, and the speech frame in the voice segments can be the full-speed voice frame, also can be the several layers of full-speed voice frame.Adopt the full-speed voice frame in the present embodiment, its frame structure is as shown in table 1:

Whether step S402, demoder have taken place by the switching of broadband to the arrowband according to the frame structure detection of encoding code stream, if switching has taken place, change step S403, otherwise, encoding code stream is decoded and export sound signal after the reconstruction according to normal decoding process.

If what receive is speech frame, then data length or the decode rate according to present frame can determine whether to have taken place by the switching of broadband to the arrowband, if for example the data of present frame only are layer 1 and layer 2, then the length of present frame is 160 bits (being that decode rate is 8kb/s) or 240 bits (being that decode rate is 12kb/s), present frame is the arrowband, otherwise, if it is more high-rise that the data of present frame also have except two-layer before comprising, the length that is present frame is equal to, or greater than 280 bits (being that decode rate is 14kb/s), and then present frame is the broadband.

Concrete, the bandwidth of the voice signal of judging according to present frame and former frame or former frame can detect current speech segment whether the switching of broadband to the arrowband has taken place.

Step S403, when the voice signal of the encoding code stream correspondence that receives by the broadband during to the switching of arrowband, demoder utilizes embedded type C ELP decoding to the low strap coding parameter that receives, and decodes the low strap component of signal

The coding parameter of the high-band component of signal that step S404, utilization receive before switching is with the low strap component of signal

Expand the high-band component of signal

Concrete, after demoder receives the speech frame of high-band coding parameter, the TDBWE coding parameter (comprising temporal envelope and frequency domain envelope) of M the speech frame that receives before each buffer memory switches, detecting by the broadband after the switching of arrowband, demoder is at first according to the temporal envelope and the frequency domain envelope of the speech frame that receives before the switching of storing in the buffer area, extrapolation goes out the temporal envelope and the frequency domain envelope of present frame, utilizes temporal envelope that extrapolation goes out and frequency domain envelope to carry out the TDBWE decoding then and can expand the high-band component of signal.The TDAC coding parameter (being the MDCT coefficient) of M the speech frame that demoder receives before also can buffer memory switching, extrapolation goes out the MDCT coefficient of present frame, and the MDCT coefficient that utilizes extrapolation to go out then carries out TDAC and decodes and can expand the high-band component of signal.

Detecting by the broadband after the switching of arrowband, speech frame for disappearance high-band coding parameter, adopt the mirror image method of interpolation to estimate the synthetic parameters of high-band component of signal, promptly the high-band coding parameter with nearest M speech frame of buffer memory in the buffer area is an image source, from the current speech frame, carry out piecewise linear interpolation, the formula of segmented line linear interpolation is:

P wherein _kThe synthetic parameters of k the speech frame high-band component of signal that expression begins to reconstruct from switching position, k=0 wherein ..., N-1, the lasting number of speech frames of N for gradually going out to handle; P _-iThe high-band coding parameter of i speech frame reception before the switching position that expression is stored in the buffer area, i=1 wherein ..., M, M is for gradually going out to handle the frame number that needs buffer memory; (a) mod (b) expression a gets surplus operation to b;

Expression rounds operation downwards.The described implementation procedure effect of formula (1) is to utilize the high-band coding parameter of preceding M the speech frame of the switching of buffer memory to estimate the high-band coding parameter that switches N the speech frame in back, reconstruct the high-band component of signal of switching N the speech frame in back by TDBWE or TDAC decoding algorithm, according to the demand of practical application, M can be the arbitrary value less than N.

Step S405, high-band component of signal to expanding Carry out the time domain shaping, the high-band component of signal after obtaining handling

Concrete, when carrying out the time domain shaping, the gain factor g (k) that becomes in the time of can introducing, this time variable factor change curve as shown in Figure 5, the time-varying gain factor is the curve of linear attenuation on log-domain.For k speech frame after switching, the high-band component of signal that expands be multiply by this time-varying gain factor, as shown in Equation (2):

${\hat{s}}_{\mathrm{HB}}^{\mathrm{ts}}\left(n\right)=g\left(k\right)\·{\hat{s}}_{\mathrm{HB}}\left(n\right)---\left(2\right)$

N=0 wherein ..., L-1 k=0 ..., N-1, L represents frame length.

Step S406, alternatively, the high-band component of signal after can adopting the time varying filter method to the time domain shaping

Carry out frequency-domain shaping, obtain the high-band component of signal after the frequency-domain shaping

Concrete, will be through the high-band component of signal of time domain shaping

By time varying filter, the frequency band that makes the high-band component of signal is along with the time slowly narrows down.Become the second order butterworth filter when employed time varying filter is in the present embodiment, be fixed as-1 its zero point, limit is in continuous variation, Fig. 6 for the time become second order butterworth filter limit and change synoptic diagram, the time varying filter limit is moved along clockwise direction, that is to say that filter transmission band will constantly reduce until 0.

When decoder processes 14kb/s or higher voice signal, wide band narrow band switching mark position fad_out_flag puts 0, filtering credit counter fad_out_count puts 0, when from a certain moment, when demoder begins to handle the voice signal of 8kb/s or 12kb/s, switching mark position, broadband, arrowband fad_out_flag puts 1, and the startup time varying filter begins the high-band component of signal of rebuilding is carried out filtering, count when filtering and to continue to carry out time-variable filtering when fad_out_count satisfies fad_out_count＜FAD_OUT_COUNT_MAX condition, otherwise stop the time-variable filtering process, wherein FAD_OUT_COUNT_MAX=N * L is transition count (for example FAD_OUT_COUNT_MAX=8000).

If i constantly, an accurate limit of time varying filter is rel (i)+img (i) * j; In the m moment, this limit accurately moves to rel (m)+img (m) * j.If interpolation is counted and is N, then k interpolation result constantly is therebetween:

rel(k)＝rel(i)×(N-k)/N+rel(m)×k/N

img(k)＝img(i)×(N-k)/N+img(m)×k/N

Can recover k filter coefficient constantly by the interpolation limit, obtain transfer function:

$H\left(z\right)=\frac{1+{2z}^{-1}+z^{-2}}{1-2\mathrm{rel}\left(k\right)z^{-1}+[{\mathrm{rel}}^2\left(k\right)+{\mathrm{img}}^2\left(k\right)]z^{-2}}$

When demoder received wideband speech signal, filtering credit counter fad_out_count was changed to 0, and the voice signal that receives when demoder is when the broadband switches to the arrowband, and time varying filter begins to start, and the filtering counter upgrades by following formula:

fad_out_count＝min(fad_out_count+1，FAD_OUT_COUNT_MAX)

Wherein FAD_OUT_COUNT_MAX is that transition period continues sampling number.

If a ₁=2rel (k), a ₂=-[rel ²(k)+img ²(k)], the reconstruction high-band component of signal of process time domain shaping

Be the input signal of time varying filter,

Output signal for time varying filter then has:

${\hat{s}}_{\mathrm{HB}}^{\mathrm{fad}}\left(n\right)=\mathrm{gain}\_\mathrm{filter}\×[a_1\×{\hat{s}}_{\mathrm{HB}}^{\mathrm{fad}}(n-1)+a_2\×{\hat{s}}_{\mathrm{HB}}^{\mathrm{fad}}(n-2)+{\hat{s}}_{\mathrm{HB}}^{\mathrm{ts}}\left(n\right)+2.0\×{\hat{s}}_{\mathrm{HB}}^{\mathrm{ts}}(n-1)+{\hat{s}}_{\mathrm{HB}}^{\mathrm{ts}}(n-2)]$

Wherein, gain_filter is a filter gain, and computing formula is:

$\mathrm{gain}\_\mathrm{filter}=\frac{1-a_1-a_2}{4}$

Step S407, utilize the QMF bank of filters with the decoding the low strap component of signal

With the high-band component of signal after the processing

If (execution in step S406 then be the high-band component of signal not

) carry out synthetic filtering, become the signal that gradually goes out in the time of can reconstructing, satisfy the characteristic that is smoothly transitted into the arrowband from the broadband.

Will be through the high-band component of signal after out-of-date change gradually goes out to handle

With the low strap component of signal that reconstructs Together, input QMF bank of filters is carried out synthetic filtering, obtains the reconstruction signal of full range band, even the switching of frequent broadband to the arrowband occur during decoding, the reconstruction signal after handling through the present invention provide acoustical quality relatively preferably still can for people's ear.

In the present embodiment, time change with voice segments gradually goes out to handle employing method one, promptly utilize the time domain gain factor that the high-band information that expands is carried out the time domain shaping, and present invention is described to utilize time-variable filtering to carry out frequency-domain shaping to the high-band information after the time domain shaping, be understandable that time becomes gradually to go out to handle and can also adopt additive method.In the embodiments of the invention three, the encoding code stream that receives with demoder is a voice segments, and in time, becomes and gradually to go out to handle employing method three, promptly utilize frequency domain high-band parameter time varying method of weighting that the high-band information that expands is carried out frequency-domain shaping and be example, a kind of coding/decoding method of sound signal as shown in Figure 7, concrete steps are as follows:

Step S401～step S403 is consistent among step S701～step S703 and the embodiment two, does not repeat them here.

The coding parameter of the high-band component of signal that step S704, utilization receive before switching is with the low strap component of signal

Expand the high-band coding parameter.

This process is to utilize the high-band coding parameter of preceding M the speech frame of switching of decoder buffer to estimate the high-band coding parameter (frequency domain envelope and high-band spectrum envelope) that switches N the speech frame in back, be specially: after demoder receives the frame that comprises the high-band coding parameter, the TDBWE coding parameter of M the speech frame that receives before each buffer memory switches, comprise temporal envelope and frequency domain envelope coding parameter, detecting the broadband after the switching of arrowband, demoder is at first according to temporal envelope that receives before the switching of storing in the buffer area and frequency domain envelope, and extrapolation goes out the temporal envelope and the frequency domain envelope of present frame.The TDAC coding parameter (being the MDCT coefficient) of M the speech frame that demoder receives before also can buffer memory switching expands the high-band coding parameter according to the MDCT coefficient of speech frame.

Detecting the broadband after the switching of arrowband, frame for disappearance high-band coding parameter, adopt the mirror image method of interpolation to estimate the synthetic parameters of high-band component of signal, promptly the high-band coding parameter (frequency domain envelope and high-band spectrum envelope) with the nearest M of buffer memory in the buffer area (for example M=5) speech frame is an image source, from the current speech frame, carry out piecewise linear interpolation, can adopt the piecewise linear interpolation formula (1) among the embodiment two to realize, wherein continuing frame number is N (for example N=50), and the effect of this process is to utilize the high-band coding parameter of preceding M the frame of the switching of buffer memory to estimate the high-band coding parameter (frequency domain envelope and high-band spectrum envelope) that switches N the frame in back.

Step S705, employing frequency domain high-band parameter time varying weighted method are carried out frequency-domain shaping to the high-band coding parameter that expands.

Concrete, high band signal is divided a plurality of subbands from frequency domain, again each subband high-band coding parameter is carried out frequency domain weighting by different gains, the frequency band of high-band component of signal is slowly narrowed down.Because no matter the coding parameter in broadband is frequency domain envelope in the TDBWE encryption algorithm when being 14kb/s or the high-band envelope in the TDAC encryption algorithm of the above speed of 14kb/s, all implying the process that high-band is divided into the subband of some, therefore, if directly the high-band coding parameter that receives is become gradually to go out to handle when frequency domain carries out, will be than on time domain, using the method for filtering to save a lot of calculated amount.When the voice signal of decoder processes 14kb/s or higher rate, switching mark position, broadband, arrowband fad_out_flag is changed to 0, transition frames counter fad_out_frame_count puts 0, when from a certain moment, when demoder begins to handle the voice signal of 8kb/s, 12kb/s, switching mark position, broadband, arrowband fad_out_flag is for being changed to 1, just the coding parameter to frequency domain is weighted when transition frame number fad_out_frame_count satisfies fad_out_frame_count＜N condition simultaneously, and weighting factor changes in time.

When speech frame rates is higher than 14kb/s before if switch, then receive and the coding parameter of the high-band component of signal of buffer memory to the buffer area comprises the high-band envelope in MDCT territory and the frequency domain envelope in the TDBWE algorithm, otherwise receive and buffer memory to the high-band signal encoding parameter in the buffer area only comprises frequency domain envelope in the TDBWE algorithm.To k speech frame after switching (k=1 ..., N), the high-band coding parameter in the use buffer area reconstructs the corresponding high-band coding parameter of present frame, the high-band envelope in frequency domain envelope or MDCT territory.Envelope on these frequency domains is divided into a plurality of subbands with whole high-band, and these spectrum envelopes are used (j=0 ..., J-1, the sub band number of J for dividing, for example G.729.1 in J=12 the frequency domain envelope in the TDBWE algorithm, J=18 for the high-band envelope in MDCT territory) expression, each subband is become the gain factor gain that gradually goes out on time, and (k j) is weighted, promptly

Become the spectrum envelope that gradually goes out in the time of can obtaining on the frequency domain.Gain (k, computing formula j) is:

$\mathrm{gain}(k,j)=\frac{\max (0,(J-j)\×N-J\×k)}{J\×N},$ k＝1，…，N?j＝0，…，J-1

To the TDBWE frequency domain envelope after handling and the high-band envelope in MDCT territory, use TDBWE decoding algorithm and the TDAC decoding algorithm high-band component of signal that change gradually goes out in the time of can obtaining of decoding respectively

Step S706, the high-band component of signal after utilizing the QMF bank of filters to handle

With the low strap component of signal that decodes

Carry out synthetic filtering, become the signal that gradually goes out in the time of can reconstructing.

Sound signal comprises voice signal and noise signal, and among embodiment two and the embodiment three, present invention is described by the example that switches to of broadband to the arrowband with voice segments, is understandable that the noise section also may be switched to the arrowband by the broadband.In the embodiment of the invention four, the encoding code stream that receives with demoder is the noise section, and in time, becomes and gradually to go out to handle to adopt method two, promptly utilize time-variable filtering to carry out frequency-domain shaping to the high-band information that expands, further can utilize the time domain gain factor to carry out time domain and be shaped as example the high-band information after the frequency-domain shaping, a kind of coding/decoding method of sound signal as shown in Figure 8, concrete steps are as follows:

Step S801, demoder receive the encoding code stream that scrambler sends, and determine the frame structure of the encoding code stream that receives.

Concrete, scrambler adopts the flow process of system shown in Figure 1 block diagram to coding audio signal, encoding code stream is sent to demoder, demoder receives encoding code stream, if encoding code stream corresponding audio signal does not take place by the switching of broadband to the arrowband, demoder adopts the flow process of system chart shown in Figure 2 that the encoding code stream that receives is carried out normal decoder, does not repeat them here.The encoding code stream that demoder receives is a noise section, and the noise frame in the noise section can be the full rate noise frame, also can be the several layers of full rate noise frame.Noise frame can be continuous programming code and transmission, it also can be the technology that adopts discontinuous transmission, noise section in the present embodiment is used identical definition with noise frame, and the noise frame that demoder receives in the present embodiment is the full rate noise frame, and the noise frame coding structure that uses in the present embodiment is as shown in table 2.

Whether step S802, demoder have taken place by the switching of broadband to the arrowband according to the frame structure detection of encoding code stream, if switching has taken place, change step S803, otherwise, encoding code stream is decoded and export noise signal after the reconstruction according to normal decoding process.

If what receive is noise frame, demoder then can determine whether to have taken place by the switching of broadband to the arrowband according to the data length of present frame, if for example the data of present frame only are narrowband core layer or narrowband core layer+arrowband enhancement layer, when the length that is present frame is 15 bits or 24 bits, then present frame is the arrowband, otherwise, if the data of present frame also comprise the broadband core layer, the length that is present frame is 43 bits, and then present frame is the broadband.

The bandwidth of the noise signal of judging according to present frame and former frame or former frame can detect the current switching of broadband to the arrowband that whether taken place.

If comprise high-band coding parameter (being the broadband core layer) in the SID that demoder receives (Silence Insertion Descriptor, the quiet insertion frame) frame, then the high-band coding parameter in the buffer area upgraded with this SID frame.From a certain moment of noise section, no longer contain the broadband core layer in the SID frame that demoder receives, then the demoder judgement has taken place by the switching of broadband to the arrowband.

Step S803, when the noise signal of the encoding code stream correspondence that receives by the broadband during to the switching of arrowband, demoder utilizes embedded type C ELP decoding to the low strap coding parameter that receives, and decodes the low strap component of signal

The coding parameter of the high-band component of signal that step S804, utilization receive before switching is with the low strap component of signal Expand the high-band component of signal

For the noise frame of disappearance high-band coding parameter, adopt the mirror image method of interpolation to estimate the synthetic parameters of high-band component of signal.If noise frame adopts the continuous programming code transmission technology, then the high-band coding parameter (frequency domain envelope and high-band spectrum envelope) with nearest M the noise frame (for example M=5) of buffer memory in the buffer area is an image source, utilizes the formula (1) among the embodiment two to be rebuild by broadband high-band coding parameter of k noise frame after the switching of arrowband having taken place; If noise frame adopts the technology of discontinuous transmission, then the SID frame that comprises high-band coding parameter (frequency domain envelope) with nearest 2 noise frame of buffer memory in the buffer area is an image source, from present frame, carries out piecewise linear interpolation.Rebuild broadband k vertical frame dimension band coding parameter after the switching of arrowband takes place with formula (3):

$P_k=\frac{k}{N-1}{P}_{\mathrm{sid}\_\mathrm{past}}+(1-\frac{k}{N-1})P_{\mathrm{sid}\_p\_\mathrm{past}}---\left(3\right)$

Lasting frame number is N (for example N=50), wherein P _{Sid_past}The high-band coding parameter of the SID frame of storing in the expression buffer area that comprises the broadband core layer recently, P _{Sid_p_past}The inferior high-band coding parameter of storing in the expression buffer area that closely comprises the SID frame of broadband core layer.This process utilizes the high-band coding parameter of preceding 2 noise frame of the switching of buffer memory to estimate the high-band coding parameter (frequency domain envelope) that switches N the noise frame in back, to recover the high-band component of signal of switching N the noise frame in back.The high-band coding parameter that utilizes formula (3) to reconstruct expands the high-band component of signal by TDBWE or TDAC decoding

Step S805, the high-band component of signal of employing time-variable filtering method to expanding Carry out frequency-domain shaping, obtain the high-band component of signal after the frequency-domain shaping

Concrete, when carrying out frequency-domain shaping, with the high-band component of signal that expands By time varying filter, the frequency band that makes the high-band component of signal is along with the time slowly narrows down, and the limit change curve of this wave filter as shown in Figure 6.When each demoder receives the SID frame that contains the broadband core layer, wide band narrow band switching mark position fad_out_flag puts 0, filtering credit counter fad_out_flag puts 0, when from a certain moment, when demoder receives the SID frame that does not comprise the broadband core layer, switching mark position, broadband, arrowband fad_out_flag puts 1, and the startup time varying filter begins the high-band component of signal of rebuilding is carried out filtering, count when filtering and to continue to carry out time-variable filtering when fad_out_count satisfies fad_out_count＜FAD_OUT_COUNT_MAX condition, otherwise stop the time-variable filtering process, wherein FAD_OUT_COUNT_MAX=N * L is transition count (for example FAD_OUT_COUNT_MAX=8000).

If i constantly, an accurate limit of time varying filter is rel (i)+img (i) * j; In the m moment, this limit accurately moves to rel (m)+img (m) * j.If interpolation is counted and is N, then k interpolation result constantly is therebetween:

rel(k)＝rel(i)×(N-k)/N+rel(m)×k/N

img(k)＝img(i)×(N-k)/N+img(m)×k/N

Can recover k filter coefficient constantly by the interpolation limit, obtain transfer function:

$H\left(z\right)=\frac{1+{2z}^{-1}+z^{-2}}{1-2\mathrm{rel}\left(k\right)z^{-1}+[{\mathrm{rel}}^2\left(k\right)+{\mathrm{img}}^2\left(k\right)]z^{-2}}$

When demoder received the BROADBAND NOISE signal, filtering counter fad_out_count was changed to 0, and when the noise signal that receives when demoder switched to the arrowband by the broadband, time varying filter began to start, and the filtering counter upgrades by following formula:

fad_out_count＝min(fad_out_count+1，FAD_OUT_COUNT_MAX)

Wherein FAD_OUT_COUNT_MAX is that transition period continues sampling number.

If a ₁=2rel (k), a ₂=-[rel ²(k)+img ²(k)], the high-band component of signal that expands

Be the input signal of time varying filter,

Output signal for time varying filter then has:

${\hat{s}}_{\mathrm{HB}}^{\mathrm{fad}}\left(n\right)=\mathrm{gain}\_\mathrm{filter}\×[a_1\×{\hat{s}}_{\mathrm{HB}}^{\mathrm{fad}}(n-1)+a_2\×{\hat{s}}_{\mathrm{HB}}^{\mathrm{fad}}(n-2)+{\hat{s}}_{\mathrm{HB}}\left(n\right)+2.0\×{\hat{s}}_{\mathrm{HB}}(n-1)+{\hat{s}}_{\mathrm{HB}}(n-2)]$

Wherein, gain_filter is a filter gain, and computing formula is:

$\mathrm{gain}\_\mathrm{filter}=\frac{1-a_1-a_2}{4}$

Step S806, alternatively can be to the high-band component of signal after the frequency-domain shaping Carry out the time domain shaping, obtain the high-band component of signal after the time domain shaping

Concrete, the gain factor g (k) that becomes when the time domain shaping can be introduced, this time variable factor change curve as shown in Figure 5.For k frame after switching, to multiply by the time-varying gain factor through the high-band component of signal that expands after TDBWE or the TDAC decoding, as shown in Equation (2), it is consistent among its implementation procedure and the embodiment two the high-band component of signal to be carried out the process of time domain shaping, does not repeat them here.Also the time gain factor that becomes in this step can be taken on the filter gain in step S805, these two kinds of methods will obtain same result.

Step S807, utilize the QMF bank of filters with the low strap component of signal that decodes

With the high-band component of signal after the shaping

If (execution in step S806 then be the high-band component of signal not

) carry out synthetic filtering, become the signal that gradually goes out in the time of can reconstructing, satisfy the characteristic that is smoothly transitted into the arrowband from the broadband.

In the present embodiment, time change with the noise section gradually goes out to handle to adopt method two, promptly utilize time-variable filtering to carry out frequency-domain shaping to the high-band information that expands, further can to the high-band information after the frequency-domain shaping utilize the time domain gain factor carry out time domain be shaped as the example present invention is described, be understandable that time becomes gradually to go out to handle and can also adopt additive method.In the embodiments of the invention five, the encoding code stream that receives with demoder is the noise section, and in time, becomes and gradually to go out to handle employing method four, promptly the high-band information that expands is divided subband, each sub-band coding parameter is carried out frequency domain high-band parameter time varying be weighted to example, a kind of method of audio decoder as shown in Figure 9, concrete steps are as follows:

Step S801～step S803 is consistent among step S901～step S903 and the embodiment four, does not repeat them here.

The coding parameter (including but not limited to the frequency domain envelope) of the high-band component of signal that step S904, utilization receive before switching expands the high-band coding parameter.

For the noise frame of disappearance high-band coding parameter, adopt the mirror image method of interpolation to estimate the synthetic parameters of high-band component of signal.If noise frame adopts the continuous programming code transmission technology, then the high-band coding parameter (frequency domain envelope and high-band spectrum envelope) with nearest M the frame (for example M=5) of buffer memory in the buffer area is an image source, can utilize formula (1) to rebuild broadband k vertical frame dimension band coding parameter after the switching of arrowband takes place; If noise frame adopts the technology of discontinuous transmission, then the SID frame that comprises high-band coding parameter (frequency domain envelope) with nearest 2 frames of buffer memory in the buffer area is an image source, from present frame, carry out piecewise linear interpolation, can utilize formula formula (3) to rebuild broadband k vertical frame dimension band coding parameter after the switching of arrowband takes place.

Because the type difference of the high-band coding parameter of different encryption algorithm sound intermediate frequency signals, the above-mentioned high-band coding parameter that expands possibly can't be divided subband, need expand the high-band component of signal this moment to the high-band coding parameter decoding that expands, the high-band component of signal that expands is extracted the high-band coding parameter again, carry out frequency-domain shaping.

Step S905, decoding expands the high-band component of signal to the high-band coding parameter that expands.

Step S906, utilize the TDBWE algorithm to extract the frequency domain envelope to the high-band component of signal that expands, these frequency domain envelopes are divided into a series of subbands that not have overlapping mutually with whole high-band component of signal.

Step S907, employing frequency domain high-band parameter time varying weighted method are carried out frequency-domain shaping to the frequency domain envelope that extracts, the high-band component of signal after decoding obtains handling to the frequency domain envelope after the frequency-domain shaping.

Concrete, when carrying out, the frequency domain envelope that utilization extracts becomes weighted, because the frequency domain envelope is equivalent to the high-band component of signal is divided a plurality of subbands from frequency domain, each frequency domain envelope is carried out frequency domain weighting by different gains, thereby the frequency band of signal is slowly narrowed down.When demoder receives the SID frame that contains the high-band coding parameter continuously, think and be in the BROADBAND NOISE signal phase at present, wide band narrow band switching mark position fad_out_flag puts 0, transition frames counter fad_out_frame_count puts 0, when from a certain moment, the SID frame that demoder receives does not comprise the broadband core layer, then demoder is thought the switching of broadband to the arrowband has been taken place, then wide band narrow band switching mark position fad_out_flag is changed to 1, when transition frame number counter fad_out_frame_count satisfies fad_out_frame_count＜N condition simultaneously, then when being carried out, the coding parameter weighting of frequency domain becomes gradually to go out to handle, and weighting factor changes in time, and wherein N is transition frame number (for example N=50).

With formula (3) to broadband k (k=0 after the switching of arrowband takes place, N-1) the high-band coding parameter of individual frame is rebuild, and the high-band coding parameter decoding that reconstructs is expanded the high-band component of signal, utilizes the TDBWE algorithm to extract the frequency domain envelope to the high-band component of signal that expands

(j=0 ..., J-1, the sub band number of J for dividing), the frequency domain envelope of each subband is become the gain factor gain that gradually goes out on time, and (k j) is weighted, promptly

Become the spectrum envelope that gradually goes out in the time of can obtaining on the frequency domain.Gain (k, computing formula j) is:

$\mathrm{gain}(k,j)=\frac{\max (0,(J-j)\×N-J\×k)}{J\×N},$ k＝1，…，N?j＝0，…，J-1

To the time after becoming the TDBWE frequency domain envelope gradually go out and using the TDBWE decoding algorithm to decode can to obtain to handle the time become the high-band component of signal that gradually goes out.

Step S908, the high-band component of signal after utilizing the QMF bank of filters to handle and the low strap component of signal that decodes

Carry out synthetic filtering, become the signal that gradually goes out in the time of can reconstructing.

The voice segments of the encoding code stream correspondence that above embodiment receives with demoder or noise section switch to example by the broadband to the arrowband, and present invention is described, be understandable that, also may have following two kinds of situations: the voice segments of the encoding code stream correspondence that demoder receives is switched to the arrowband by the broadband, and switches the noise section that the back demoder still can receive the encoding code stream correspondence; Or the noise section of the encoding code stream correspondence that receives of demoder switched to the arrowband by the broadband, and switches the voice segments that the back demoder still can receive the encoding code stream correspondence.

In the embodiment of the invention six, the voice segments of the encoding code stream correspondence that receives with demoder is switched to the arrowband by the broadband, switch the noise section that the back demoder still can receive the encoding code stream correspondence, and in time, becomes and gradually to go out to handle employing method three, promptly utilize frequency domain high-band parameter time varying method of weighting that the high-band information that expands is carried out frequency-domain shaping and be example, a kind of method of audio decoder as shown in figure 10, concrete steps are as follows:

Step S1001, demoder receive the encoding code stream that scrambler sends, and determine the frame structure of the encoding code stream that receives.

Concrete, scrambler adopts the flow process of system shown in Figure 1 block diagram to coding audio signal, encoding code stream is sent to demoder, demoder receives encoding code stream, if encoding code stream corresponding audio signal does not take place by the switching of broadband to the arrowband, demoder adopts the flow process of system chart shown in Figure 2 that the encoding code stream that receives is carried out normal decoder, does not repeat them here.The encoding code stream that demoder receives in the present embodiment comprises voice segments and noise section, and wherein the speech frame in the voice segments adopts the frame structure of the full-speed voice frame shown in the table 1, and the noise frame in the noise section adopts the full rate noise frame structure shown in the table 2.

Whether step S1002, demoder have taken place by the switching of broadband to the arrowband according to the frame structure detection of encoding code stream, if switching has taken place, change step S1003, otherwise, encoding code stream is decoded and export sound signal after the reconstruction according to normal decoding process.

Step S1003, when the voice signal of the encoding code stream correspondence that receives by the broadband during to the switching of arrowband, demoder utilizes embedded type C ELP decoding to the low strap coding parameter that receives, and decodes the low strap component of signal

Step S1004, utilize artificial band spreading technique, with the low strap component of signal

Expand the high-band coding parameter.

When taking place the broadband to the switching of arrowband, the sound signal of storing in the buffer area may be identical or different with the type of the sound signal that receives after the switching, may have following five kinds of situations:

What (1) store in the buffer area all is the high-band coding parameter (promptly have only the frequency domain envelope of TDBWE, and do not have the high-band envelope of TDAC) of noise frame, and the frame that receives after switching all is a speech frame;

What (2) store in the buffer area all is the high-band coding parameter (promptly have only the frequency domain envelope of TDBWE, and do not have the high-band envelope of TDAC) of noise frame, and the frame that receives after switching all is a noise frame;

What (3) store in the buffer area all is the high-band coding parameter (the frequency domain envelope of existing TDBWE also has the high-band envelope of TDAC) of speech frame, and the frame that receives after switching all is a speech frame;

What (5) store in the buffer area all is the high-band coding parameter (the frequency domain envelope of existing TDBWE also has the high-band envelope of TDAC) of speech frame, and the frame that receives after switching all is a noise frame;

(6) the high-band coding parameter of the existing speech frame of storing in the buffer area (the frequency domain envelope of existing TDBWE, the high-band envelope that TDAC is also arranged), high-band coding parameter (the frequency domain envelope that has only TDBWE that noise frame is also arranged, and do not have the high-band envelope of TDAC), the existing noise frame of the frame that receives after switching also has speech frame.

Wherein situation (2) and situation (3) are described in detail in the above-described embodiments, for remaining three kinds of situation, after switching, all can rebuild the high-band coding parameter according to the method for formula (1).But owing to there is not the high-band envelope of TDAC in the high-band coding parameter of noise frame, therefore, for the situation that but receives the noise section after switching in voice segments, no longer rebuild the high-band coding parameter, promptly all no longer rebuild the high-band envelope of TDAC, because just to a kind of enhancing of TDBWE coding, only the frequency domain envelope with TDBWE can recover the high-band component of signal to the TDAC encryption algorithm fully.In other words, be exactly the stage that starts the present embodiment scheme (promptly switch and take place in the N frame of back), be that 14kb/s decodes for the equal reduction of speed of speech frame, until finish become when whole gradually to go out to operate after.For the k frame after switching (k=1 ..., N), reconstruct the frequency domain envelope of high-band coding parameter

(j=0 ..., J-1, J=12).

Step S1005, employing frequency domain high-band parameter time varying method of weighting are carried out frequency-domain shaping to the high-band coding parameter that expands, to the decoding of the high-band coding parameter after the shaping, and the high-band component of signal after obtaining to handle.

Concrete, when frequency-domain shaping, high band signal is divided a plurality of subbands from frequency domain, again each subband or the high-band coding parameter that characterizes each subband are carried out frequency domain weighting by different gains, the frequency band of signal is slowly narrowed down.Frequency domain envelope in the TDBWE encryption algorithm that uses in the speech frame or for the frequency domain envelope in the core layer of noise frame broadband is all implying the process that high-band is divided into the subband of some.Demoder receives the sound signal that contains the high-band coding parameter (comprising the SID frame that contains the broadband core layer and the speech frame of 14kb/s and higher rate), wide band narrow band switching mark position fad_out_flag puts 0, transition frames counter fad_out_frame_count puts 0, when from a certain moment, the sound signal that demoder receives does not comprise high-band coding parameter (being the speech frame that does not have the broadband core layer in the SID frame or be lower than 14kb/s), then demoder is thought the switching of broadband to the arrowband has been taken place, then wide band narrow band switching mark position fad_out_flag is changed to 1, when transition frame number counter fad_out_frame_count satisfies fad_out_frame_count＜N condition simultaneously, then when being carried out, the coding parameter weighting of frequency domain becomes gradually to go out to handle, and weighting factor changes in time, and wherein N is transition frame number (for example N=50).

J frequency domain envelope is divided into J subband with the high-band component of signal, and (k j) is weighted the gain factor gain that each frequency domain envelope is become on time, promptly

Become the spectrum envelope gradually go out in the time of can obtaining on the frequency domain, wherein gain (k, computing formula j) is:

$\mathrm{gain}(k,j)=\frac{\max (0,(J-j)\×N-J\×k)}{J\×N},$ k＝1，…，N?j＝0，…，J-1

After TDBWE frequency domain envelope after handling used the TDBWE decoding algorithm to decode can to obtain to handle the time become the high-band component of signal that gradually goes out.

Step S1006, the high-band component of signal after utilizing the QMF bank of filters to handle and the low strap component of signal that decodes

Carry out synthetic filtering, become the signal that gradually goes out in the time of can reconstructing.

In the embodiment of the invention seven, the noise section of the encoding code stream correspondence that receives with demoder is switched to the arrowband by the broadband, switch the voice segments that the back demoder still can receive the encoding code stream correspondence, and in time, becomes and gradually to go out to handle employing method three, promptly utilize frequency domain high-band parameter time varying method of weighting that the high-band information that expands is carried out frequency-domain shaping and be example, a kind of method of audio decoder as shown in figure 11, concrete steps are as follows:

Step S1001～step S1002 is consistent among step S1101～step S1102 and the embodiment six, does not repeat them here.

Step S1103, when the noise signal of the encoding code stream correspondence that receives by the broadband during to the switching of arrowband, demoder utilizes embedded type C ELP decoding to the low strap coding parameter that receives, and decodes the low strap component of signal

Step S1104, utilize artificial band spreading technique, with the low strap component of signal

Expand the high-band coding parameter.

Step S1105, employing frequency domain high-band parameter time varying method of weighting are carried out frequency-domain shaping to the high-band coding parameter that expands, to the decoding of the high-band coding parameter after the shaping, and the high-band component of signal after obtaining to handle.

Concrete, when frequency-domain shaping,, the frequency band of signal is slowly broadened with representing the high-band coding parameter of each subband to carry out frequency domain weighting by different gains.Demoder receives the sound signal that contains the wideband encoding parameter (comprising the SID frame that contains the broadband core layer and the speech frame of 14kb/s and higher rate), wide band narrow band switching mark position fad_out_flag puts 0, transition frames counter fad_out_frame_count puts 0, when from a certain moment, the sound signal that demoder receives does not comprise wideband encoding parameter (being the speech frame that does not have the broadband core layer in the SID frame or be lower than 14kb/s), then demoder is thought the switching of broadband to the arrowband has been taken place, then wide band narrow band switching mark position fad_out_flag is changed to 1, when transition frame number counter fad_out_frame_count satisfies fad_out_frame_count＜N condition simultaneously, then when being carried out, the coding parameter weighting of frequency domain becomes gradually to go out to handle, and weighting factor changes in time, and wherein N is transition frame number (for example N=50).

In the present embodiment, when switching, what store in the buffer area all is the wideband encoding parameter (promptly have only the frequency domain envelope of TDBWE, and do not have the high-band envelope of TDAC) of noise frame, and the existing noise frame of the frame of receiving after switching also has speech frame.After switching, need rebuild the interior high-band coding parameter of scheme duration of implementing present embodiment according to the method for formula (1), but high-band envelope parameters owing to the TDAC that need in the speech frame not to have in the high-band coding parameter of noise, therefore for the speech frame that receives when rebuilding the high-band coding parameter, all no longer rebuild the high-band envelope of TDAC, because just to a kind of enhancing of TDBWE coding, only the frequency domain envelope with TDBWE can recover the high-band component of signal to the TDAC encryption algorithm fully.In other words, be exactly the stage that starts the present embodiment scheme (promptly switch and take place in the N frame of back), be that 14kb/s decodes for the equal reduction of speed of speech frame, until finish become when whole gradually to go out to operate after.For the k frame after switching (k=1 ..., N), the high wideband encoding parameter that reconstructs is the frequency domain envelope

(j=0 ..., J-1 J=12) is divided into the high-band component J subband, and each subband is become the gain factor gain that gradually goes out on time, and (k j) is weighted, promptly

Become the spectrum envelope gradually go out in the time of can obtaining on the frequency domain, wherein gain (k, computing formula j) is:

$\mathrm{gain}(k,j)=\frac{\max (0,(J-j)\×N-J\×k)}{J\×N},$ k＝1，…，N?j＝0，…，J-1

When being used the TDBWE decoding algorithm to decode can to obtain, the TDBWE frequency domain envelope after handling becomes the high-band component of signal that gradually goes out.

Step S1106, the high-band component of signal after utilizing the QMF bank of filters to handle and the narrow band signal component that decodes

Carry out synthetic filtering, become the signal that gradually goes out in the time of can reconstructing.

In the embodiment of the invention eight, the voice segments of the encoding code stream correspondence that receives with demoder is switched to the arrowband by the broadband, switch the noise section that the back demoder still can receive the encoding code stream correspondence, and the time become the short-cut method gradually go out to handle employing method three, a kind of method of audio decoder as shown in figure 12, concrete steps are as follows:

Step S1001～step S1002 is consistent among step S1201～step S1202 and the embodiment six, does not repeat them here.

Step S1203, when the voice signal that receives by the broadband during to the switching of arrowband, demoder utilizes embedded type C ELP decoding to the low strap coding parameter that receives, and decodes the low strap component of signal

Step S1204, utilize artificial band spreading technique, with the low strap component of signal

Expand the high-band coding parameter.

When taking place the broadband to the switching of arrowband, the sound signal of storing in the buffer area may with switch after the type of the sound signal that receives identical or different, comprise five kinds of situations described in embodiment six, wherein situation (2) and situation (3), be described in detail in the above-described embodiments, for remaining three kinds of situation, after switching, all can rebuild the high-band coding parameter according to the method for formula (1).But, therefore, all no longer rebuild the high-band envelope of TDAC, and just rebuild the frequency domain envelope in the TDBWE algorithm for the coding parameter of rebuilding owing to there is not the high-band envelope of TDAC in the high-band coding parameter of noise

Because just to a kind of enhancing of TDBWE coding, only the frequency domain envelope with TDBWE can recover the high-band component of signal to the TDAC encryption algorithm fully.In other words, be exactly (promptly to switch back COUNT takes place in the stage that starts the present invention program _{Fad_out}In the frame), be that 14kb/s decodes for the equal reduction of speed of speech frame, until finish become when whole gradually go out operation after.For the k frame after switching (k=0 ..., COUNT _{Fad_out}-1) the high-band coding parameter that, reconstructs is the frequency domain envelope (j=0 ..., J-1) the high-band component of signal is divided into J subband.

The method that step S1205, utilization are simplified is carried out frequency-domain shaping to the high-band coding parameter that expands, to the decoding of the high-band coding parameter after the shaping, and the high-band component of signal after obtaining to handle.

When frequency-domain shaping, the frequency domain envelope that reconstructs

High band signal is divided J subband from frequency domain.When wide band narrow band switching mark position fad_out_flag is 1, and transition frames counter fad_out_frame_count satisfies fad_out_frame_count＜COUNT simultaneously _{Fad_out}During condition, then gradually go out processing for switching to become when frequency domain envelope that k the frame in back reconstruct carries out according to formula (4) or formula (5) or formula (6):

Wherein

Expression is no more than the maximum integer of x.Become the high-band component of signal that gradually goes out when using the TDBWE decoding algorithm to obtain to the TDBWE frequency domain envelope after handling.LOW_LEVEL is frequency domain envelope possible minimum value in quantization table, for example: the frequency domain envelope (j=0 ... 3) adopt the scalar quantization technology, first order code book is:

The second level quantizes code book:

Then ${\hat{F}}_{\mathrm{env}}\left(j\right)=l1\left(j\right)+l2\left(j\right),$ Wherein l1 (j) is the quantization vector of the first order, and l2 (j) is partial quantization vector.Therefore, in the present embodiment,

Minimum value be-3.0000+ (12.95541)=-15.95541.Further, when reality was used, this minimum value can be reduced to selected an enough little value to get final product.

But need further be pointed out that, above-mentioned

Definite method be the preferred embodiments of the present invention, in actual applications, can be according to Technology Need, numerical value simplified or replaced with other numerical value that meet technical requirement, above variation belongs to protection scope of the present invention equally.

Step S1206, the high-band component of signal after utilizing the QMF bank of filters to handle and the low strap component of signal that goes out of decoding and rebuilding carry out synthetic filtering, become the signal that gradually goes out in the time of can reconstructing.

The present invention is not only applicable to broad and takes switching than the arrowband to, also be applicable to the switching of ultra broadband to the broadband, among the embodiment described above, adopt TDBWE or TDAC decoding algorithm to decode to the high-band component of signal, need to prove that the present invention also is applicable to other wideband encoding algorithms except TDBWE and TDAC decoding algorithm; In addition, the method to switch back expansion high-band component of signal and high-band coding parameter also has different implementation methods, does not repeat them here.

The method that provides by the embodiment of the invention, when sound signal taken place the broadband to the switching of arrowband the time, can utilize bandwidth detection, artificial band spread, time to become gradually to go out to handle and a series of processing such as bandwidth is synthetic, can carry out the transition to narrow band signal from broadband signal smoothly to realize switching, thereby make people's ear obtain comparatively comfortable auditory perception.

In the embodiments of the invention nine, a kind of device of audio decoder comprises as shown in figure 12:

Acquiring unit 10, be used for when the encoding code stream corresponding audio signal that receives by wider bandwidth when switching than narrow bandwidth, obtain the low strap component of signal of sound signal, and send to expanding element 20.Expanding element 20 is used for the low strap component of signal is expanded high-band information, and becomes when the high-band information that expands sent to and gradually go out processing unit 30.In time, becomes and gradually to go out processing unit 30, becomes gradually to go out to handle when being used for the high-band information that expands carried out, and the high-band component of signal after obtaining handling, and the high-band component of signal after will handling sends to synthesis unit 40.Synthesis unit 40, the low strap component of signal that high-band component of signal after the processing that is used for receiving and acquiring unit 10 obtain is synthesized.

This device also comprises: processing unit 50 is used to determine the frame structure of the encoding code stream that receives, and the frame structure of encoding code stream is sent to detecting unit 60.Whether detecting unit 60, the frame structure that is used for the encoding code stream that sends according to processing unit 50 detect and have taken place by wider bandwidth to the switching than narrow bandwidth, to the switching than narrow bandwidth encoding code stream are sent to acquiring unit 10 if taken place by wider bandwidth.

Concrete, expanding element 20 further comprises at least one in the following subelement: the first expansion subelement 21, the second expansion subelement 22, the 3rd expansion subelement 23.The first expansion subelement 21 is used to utilize the coding parameter of the high-band component of signal that receives before the switching that the low strap component of signal is expanded the high-band coding parameter.The second expansion subelement 22 is used to utilize the coding parameter of the high-band component of signal that receives before the switching that the low strap component of signal is expanded the high-band component of signal.The 3rd expansion subelement 23 is used to utilize the low strap component of signal that current audio frame decodes after switching to expand the high-band component of signal.

In time, become gradually to go out processing unit 30 and further comprise in the following subelement at least one: separating treatment subelement 31, hybrid processing subelement 32.Separating treatment subelement 31 be used for when the high-band information that expands is the high-band component of signal high-band component of signal that expands being carried out time domain shaping and/or frequency-domain shaping, and the high-band component of signal after will handling sends to synthesis unit 40.Hybrid processing subelement 32 is used for when the high-band information that expands is the high-band coding parameter high-band coding parameter that expands being carried out frequency-domain shaping; Maybe when the high-band information that expands is the high-band component of signal, the high-band component of signal that expands is divided subband, each sub-band coding parameter is carried out frequency-domain shaping, and the high-band component of signal after will handling sends to synthesis unit 50.

Separating treatment subelement 31 further comprises at least one in the following subelement: first subelement 311, second subelement 312, the 3rd subelement 313, the 4th subelement 314.First subelement 311 is used for utilizing the time domain gain factor to carry out the time domain shaping to the high-band component of signal that expands, and the high-band component of signal after handling is sent to synthesis unit 40.Second subelement 312 is used for utilizing time-variable filtering to carry out frequency-domain shaping to the high-band component of signal that expands, and the high-band component of signal after handling is sent to synthesis unit 40.The 3rd subelement 313, be used for utilizing the time domain gain factor to carry out the time domain shaping to the high-band component of signal that expands, and utilize time-variable filtering to carry out frequency-domain shaping to the high-band component of signal after the time domain shaping, the high-band component of signal after handling is sent to synthesis unit 40.The 4th subelement 314, be used for utilizing time-variable filtering to carry out frequency-domain shaping to the high-band component of signal that expands, and utilize the time domain gain factor to carry out the time domain shaping to the high-band component of signal after the frequency-domain shaping, the high-band component of signal after handling is sent to synthesis unit 40.

Hybrid processing subelement 32 further comprises at least one in the following subelement: the 5th subelement 321, the 6th subelement 322.The 5th subelement 321, be used for when the high-band information that expands is the high-band coding parameter, utilize frequency domain high-band parameter time varying method of weighting that the high-band coding parameter that expands is carried out frequency-domain shaping, become the spectrum envelope that gradually goes out when obtaining, decoding obtains the high-band component of signal, and the high-band component of signal after will handling sends to synthesis unit 40.The 6th subelement 322, be used for when the high-band information that expands is the high-band component of signal, the high-band component of signal that expands is divided subband, each sub-band coding parameter is carried out the weighting of frequency domain high-band parameter time varying, become the spectrum envelope that gradually goes out when obtaining, decoding obtains the high-band component of signal, and the high-band component of signal after will handling sends to synthesis unit 40.

The device that provides by the foregoing description, when sound signal taken place the broadband to the switching of arrowband the time, can utilize bandwidth detection, artificial band spread, time to become gradually to go out to handle and a series of processing such as bandwidth is synthetic, can carry out the transition to narrow band signal from broadband signal smoothly to realize switching, thereby make people's ear obtain comparatively comfortable auditory perception.

Through the above description of the embodiments, those skilled in the art can be well understood to the present invention and can realize by the mode that software adds essential general hardware platform, can certainly pass through hardware, but the former is better embodiment under a lot of situation.Based on such understanding, the part that technical scheme of the present invention contributes to prior art in essence in other words can embody with the form of software product, this obtains the machine software product and is stored in the storage medium, comprises that some instructions are used so that a station terminal equipment is carried out the described method of each embodiment of the present invention.

More than disclosed only be several specific embodiment of the present invention, still, the present invention is not limited thereto, any those skilled in the art can think variation all should fall into protection scope of the present invention.

Claims (16)

1, a kind of coding/decoding method of sound signal is characterized in that, may further comprise the steps:

When the encoding code stream corresponding audio signal that receives by wider bandwidth when switching than narrow bandwidth, obtain the low strap component of signal of described sound signal and expand high-band information;

When being carried out, the described high-band information that expands becomes gradually to go out to handle the high-band component of signal after obtaining handling;

High-band component of signal after the described processing and the described low strap component of signal of obtaining are synthesized.

2, the coding/decoding method of sound signal according to claim 1, it is characterized in that, described when the encoding code stream corresponding audio signal that receives by wider bandwidth when switching than narrow bandwidth, obtain the low strap component of signal of described sound signal and expand and also comprise before the high-band information:

Determine the frame structure of the described encoding code stream that receives, detect according to described frame structure and whether taken place by wider bandwidth to switching than narrow bandwidth.

3, the coding/decoding method of sound signal according to claim 1 is characterized in that, the described method that the low strap component of signal is expanded high-band information is specially:

Utilize the coding parameter of the high-band component of signal that receives before switching that described low strap component of signal is expanded high-band information, described high-band information is the high-band decoding parametric; Or

Utilize the coding parameter of the high-band component of signal that receives before switching that described low strap component of signal is expanded high-band information, described high-band information is the high-band component of signal; Or

Utilize the low strap component of signal that current audio frame decodes after switching to expand the high-band component of signal.

As the coding/decoding method of sound signal as described in the claim 3, it is characterized in that 4, the coding parameter of the high-band component of signal that described utilization receives before switching is specially the method that described low strap component of signal expands high-band information:

The high-band coding parameter of the audio frame that buffer memory receives before switching utilizes extrapolation to estimate the high-band coding parameter of current audio frame after the switching.

As the coding/decoding method of sound signal as described in the claim 3, it is characterized in that 5, the coding parameter of the high-band component of signal that described utilization receives before switching is specially the method that described low strap component of signal expands high-band information:

The high-band coding parameter of the audio frame that buffer memory receives before switching utilizes extrapolation to estimate the high-band coding parameter of current audio frame after the switching, the corresponding wideband decoded algorithm of high-band coding parameter utilization that described extrapolation is gone out expands the high-band component of signal.

6, the coding/decoding method of sound signal according to claim 1 is characterized in that, becomes when described high-band information is carried out gradually to go out to handle specifically to comprise: become during separation gradually to go out to handle or become gradually to go out to handle when mixing.

7, as the coding/decoding method of sound signal as described in the claim 6, it is characterized in that described high-band information is the high-band component of signal, become gradually to go out to handle during described separation and be specially:

Utilize the time domain gain factor to carry out the time domain shaping to the described high-band component of signal that expands; Or

Utilize time-variable filtering to carry out frequency-domain shaping to the described high-band component of signal that expands.

8, as the coding/decoding method of sound signal as described in the claim 7, it is characterized in that, describedly utilize the time domain gain factor to carry out also comprising after the time domain shaping the described high-band component of signal that expands:

Utilize time-variable filtering to carry out frequency-domain shaping to the high-band component of signal after the described time domain shaping.

9, as the coding/decoding method of sound signal as described in the claim 7, it is characterized in that, describedly utilize time-variable filtering to carry out also comprising after the frequency-domain shaping the described high-band component of signal that expands:

Utilize the time domain gain factor to carry out the time domain shaping to the high-band component of signal after the described frequency-domain shaping.

10, as the coding/decoding method of sound signal as described in the claim 6, it is characterized in that, become gradually to go out to handle during described mixing and be specially:

When described high-band information is the high-band coding parameter, utilize frequency domain high-band parameter time varying method of weighting that the described high-band coding parameter that expands is carried out frequency-domain shaping, become the spectrum envelope that gradually goes out when obtaining, decoding obtains the high-band component of signal; Or

When described high-band information is the high-band component of signal, the described high-band component of signal that expands is divided subband, each sub-band coding parameter is carried out the weighting of frequency domain high-band parameter time varying, become the spectrum envelope that gradually goes out when obtaining, decoding obtains the high-band component of signal.

11, a kind of decoding device of sound signal is characterized in that, comprising:

Acquiring unit, be used for when the encoding code stream corresponding audio signal that receives by wider bandwidth when switching than narrow bandwidth, obtain the low strap component of signal of described sound signal, and send to expanding element;

Expanding element is used for described low strap component of signal is expanded high-band information, and becomes when the described high-band information that expands sent to and gradually go out processing unit;

In time, becomes and gradually to go out processing unit, becomes gradually to go out to handle when being used for the described high-band information that expands carried out, and the high-band component of signal after obtaining handling, and the high-band component of signal after the described processing sent to synthesis unit;

Synthesis unit, the low strap component of signal that high-band component of signal after the described processing that is used for receiving and described acquiring unit obtain is synthesized.

12, as the decoding device of sound signal as described in the claim 11, it is characterized in that, also comprise:

Processing unit is used for determining the frame structure of the described encoding code stream that receives, and the frame structure of described encoding code stream is sent to detecting unit;

Detecting unit, whether the frame structure that is used for the described encoding code stream that sends according to described processing unit detects and has taken place by wider bandwidth to the switching than narrow bandwidth, to switching described encoding code stream is sent to described acquiring unit if taken place than narrow bandwidth by wider bandwidth.

As the decoding device of sound signal as described in the claim 11, it is characterized in that 13, described expanding element further comprises at least one in the first expansion subelement, the second expansion subelement, the 3rd expansion subelement,

The described first expansion subelement is used to utilize the coding parameter of the high-band component of signal that receives before the switching that described low strap component of signal is expanded the high-band coding parameter;

The described second expansion subelement is used to utilize the coding parameter of the high-band component of signal that receives before the switching that described low strap component of signal is expanded the high-band component of signal;

Described the 3rd expansion subelement is used to utilize the low strap component of signal that current audio frame decodes after switching to expand the high-band component of signal.

14, as the decoding device of sound signal as described in the claim 11, it is characterized in that, become when described gradually to go out processing unit and further comprise separating treatment subelement or hybrid processing subelement,

Described separating treatment subelement is used for the described high-band component of signal that expands is carried out time domain shaping and/or frequency-domain shaping, and the high-band component of signal after will handling sending to described synthesis unit when the described high-band information that expands is the high-band component of signal;

Described hybrid processing subelement is used for when the described high-band information that expands is the high-band coding parameter the described high-band coding parameter that expands being carried out frequency-domain shaping; Or when the described high-band information that expands is the high-band component of signal, the described high-band component of signal that expands is divided subband, each sub-band coding parameter is carried out frequency-domain shaping, and the high-band component of signal after will handling sends to described synthesis unit.

15, as the decoding device of sound signal as described in the claim 14, it is characterized in that described separating treatment subelement further comprises at least one in first subelement, second subelement, the 3rd subelement, the 4th subelement,

Described first subelement is used for utilizing the time domain gain factor to carry out the time domain shaping to the described high-band component of signal that expands, and the high-band component of signal after handling is sent to described synthesis unit;

Described second subelement is used for utilizing time-variable filtering to carry out frequency-domain shaping to the described high-band component of signal that expands, and the high-band component of signal after handling is sent to described synthesis unit;

Described the 3rd subelement, be used for utilizing the time domain gain factor to carry out the time domain shaping to the described high-band component of signal that expands, and utilize time-variable filtering to carry out frequency-domain shaping to the high-band component of signal after the time domain shaping, the high-band component of signal after handling is sent to described synthesis unit;

Described the 4th subelement, be used for utilizing time-variable filtering to carry out frequency-domain shaping to the described high-band component of signal that expands, and utilize the time domain gain factor to carry out the time domain shaping to the high-band component of signal after the frequency-domain shaping, the high-band component of signal after handling is sent to described synthesis unit.

16, as the decoding device of sound signal as described in the claim 14, it is characterized in that described hybrid processing subelement further comprises at least one in the 5th subelement, the 6th subelement,

Described the 5th subelement, be used for when the described high-band information that expands is the high-band coding parameter, utilize frequency domain high-band parameter time varying method of weighting that the described high-band coding parameter that expands is carried out frequency-domain shaping, become the spectrum envelope that gradually goes out when obtaining, decoding obtains the high-band component of signal, and the high-band component of signal after will handling sends to described synthesis unit;

Described the 6th subelement, be used for when the described high-band information that expands is the high-band component of signal, the described high-band component of signal that expands is divided subband, each sub-band coding parameter is carried out the weighting of frequency domain high-band parameter time varying, become the spectrum envelope that gradually goes out when obtaining, decoding obtains the high-band component of signal, and the high-band component of signal after will handling sends to described synthesis unit.

Images